Jordan Terry
TalkRL: The Reinforcement Learning Podcast
00:00
The Difference Between Truncation and Photo Finish Ending?
If you stop the simulation at some point, and the agent actually thinks that's the end of an episode, then it may just fall on its face. What you really want is to train a policy that can keep going. You can also have your turn truncated. And if you hae backwards compatibility for larning cot forthat simply just add an extra omc underscore,. It's a trivil o, changed up a learning code to handle this.
Transcript
Play full episode
